Use AI to create the next useful learning step, not to remove every step between the question and the answer.
Ask for hints, questions, explanations at a chosen level, similar practice, or feedback on your attempt. Then close the chat and demonstrate the skill independently. For graded work, follow the course’s AI policy before sharing material or using assistance.
Define the learning task first
An assignment can ask for a product, but the learning task may be reasoning, recall, design, debugging, or explanation.
Suppose the prompt asks you to write a loop. The visible product is code. The learning task may include:
- deciding what repeats;
- tracing state across iterations;
- choosing a stopping condition;
- testing boundary cases; and
- explaining why the loop terminates.
If AI supplies finished code immediately, the product may appear while the learning task remains undone.
Before prompting, write:
Skill I am practicing:
What I must do independently:
AI use allowed by the course:
Help that would preserve the next reasoning step:
How I will verify learning without AI:
When the policy is unclear, ask the instructor. Do not infer permission from the tool’s availability.
Ask for a hint, not the destination
A useful hint points to relevant evidence or a smaller question:
Do not write the solution. Give me one hint about what state must change during
each loop iteration. Wait for my response.
After using the hint, attempt the next step yourself. If still stuck, show the new attempt and ask for another limited hint.
Hints can still be wrong or too revealing. Compare them with course materials, the specification, and observable results.
Request Socratic questions
Socratic-style interaction asks the student to state assumptions and reason through consequences:
Ask me one question at a time about this algorithm. Do not provide code. Begin
by checking what I believe the input and output are. After each answer, identify
one inconsistency or ask the next question.
This does not guarantee good tutoring. The AI may ask an irrelevant question, accept a weak answer, or steer toward a different method. Keep the original learning objective visible.
OpenAI’s current Study Mode documentation describes an optional ChatGPT mode that can ask guiding questions, explain concepts in layers, and quiz the user. Those are product capabilities, not proof that every response is pedagogically or factually correct.
Ask for explanations at a specific level
“Explain recursion” gives no information about the learner or desired depth. Specify:
Explain recursion to a first-semester programming student who understands
functions and local variables but not call stacks. Use one three-call trace. Do
not use my assignment. End with one question that checks whether I can identify
the base case.
Then deepen the explanation:
- ask which detail was simplified;
- request the formal term;
- compare with the textbook definition;
- draw the trace yourself; and
- explain the concept back without copying.
Different levels should add precision, not merely more words.
Use practice that transfers
If AI solves the exact graded task, create a separate practice problem only when course policy permits:
Create a new practice problem that uses the same concept but different names,
values, and context. Do not reproduce or transform the assigned problem. Give
only the prompt. Wait for my attempt.
After attempting it, request feedback tied to criteria:
Review my explanation for these criteria: correct input and output, stated
invariant, stopping condition, and one boundary case. Point to the first missing
criterion. Do not rewrite the answer.
Feedback should direct revision. A polished replacement answer does not show what the student can do.
Preserve productive struggle
Struggle is productive when it generates information: a trace, attempted proof, test result, error message, or revised question. Repeating the same action without new evidence is not a learning requirement.
Use a stopping rule:
- attempt the next step for a defined period;
- record what you tried and observed;
- ask for the smallest useful hint;
- attempt again;
- consult course support when the obstacle persists.
AI should not replace office hours, disability accommodations, authoritative course materials, or human feedback when those are the appropriate support.
End with independent verification
Close the chat and perform one transfer check:
- solve a similar problem;
- explain the concept aloud;
- predict output before running code;
- reproduce the procedure from a blank page;
- identify and correct a new error; or
- answer a question using only permitted references.
Then compare with a primary source, official documentation, tests, or instructor feedback.
If you cannot perform the check, the interaction produced exposure rather than independent understanding. Return to the first missing step.
A four-turn tutoring sequence
For a loop problem:
- Student attempt: State input, output, loop state, and current code.
- Hint: Ask for one question about the stopping condition.
- Practice: Trace a different three-item example by hand.
- Teach-back: Explain why the loop stops and test empty input.
Keep the student’s attempt and revisions. Do not submit the AI conversation as evidence of mastery unless the course explicitly requests it.
Common mistakes
- Asking for the complete answer first. Preserve the next reasoning step.
- Using vague prompts. Name the skill, level, and help boundary.
- Confusing fluent feedback with correct feedback. Verify it.
- Practicing only the assigned example. Test transfer on a distinct case.
- Letting struggle become unproductive repetition. Seek targeted support.
- Ignoring course policy or privacy. Check before sharing material.
- Leaving understanding inside the chat. Perform an independent check.
Do this now
Choose one current concept. Rewrite an answer-seeking prompt as a request for one hint or question. Attempt the next step, then explain the concept without the chat.
Log what you learned
Record only:
- Result: What did the action produce?
- Evidence: What observation, test, or source supports that result?
- Next action or unresolved question: What should happen next?
Next, strengthen the process by writing down what you know before you ask AI.
For a longer view, use the semester practice plan for broad foundations, initiative, and mental robustness to develop capabilities that one tutoring conversation cannot supply.